Advanced AI Search & Summarization for Smoking Cessation Info

Overview

The No-smoking Guide (nosmokeguide.go.kr), operated by the Ministry of Health and Welfare and the Korea Health Promotion Institute, is South Korea's representative smoking cessation portal, providing various information and action programs. To meet the rapidly changing digital environment and rising public expectations, this project aims to implement an "Evidence-based Smoking Cessation AI Search Service" powered by Hyper-scale AI technology utilizing accumulated smoking cessation databases. This initiative is designed to enhance user convenience and public satisfaction with online smoking cessation services. Furthermore, it serves as a primary communication channel for national tobacco control policies, providing accurate health information and countering misinformation from the tobacco industry.

Project description

To train the model for this service, we collected and refined smoking cessation data accumulated since 2001, internal databases, and tobacco harm reports from the Korea Disease Control and Prevention Agency (KDCA). Through data preprocessing, we established 1,093 sets of training data and 1,039 sets of TagRules. Following procedures for smoking cessation data training and generative AI algorithm development, we built a Retrieval-Augmented Generation (RAG) based search service capable of providing customized responses to user inquiries.

Starting with a Proof of Concept (PoC) in 2023, we performed optimization of specialized search and generative AI models in 2024, successfully increasing AI search accuracy from 83% to 95%. This service provides intuitive and efficient information retrieval for citizens seeking smoking prevention and cessation information. Based on the latest internal data, the system understands user intent and summarizes core content. When users enter questions regarding tobacco control policies, laws, the risks of smoking, or effective cessation methods, the system provides immediate summaries and an "Original Document View" feature for in-depth exploration.

Key technological features

• Establishment of Training Data and Tag Rules: Constructing Tag Rules based on existing tags to enable document retrieval utilizing tag information
• Application of RAG Solution (VAIV SearchGPT): Transitioning from simple keyword-based search to semantic search (VAIV Search) to improve search accuracy and support natural language inquiries

Expected outcomes

For service commercialization, we have completed the development of both user and administrator interfaces. Since late November 2024, the "No-smoking Guide AI Search & Summary" service has been in pilot operation on the No-smoking Guide website. During its first month of operation in 2024, the service recorded 5,430 uses, achieving a high overall satisfaction score of 4.6 out of 5.0.

By moving beyond traditional keyword-based searches and applying generative AI technology, we can provide users with the smoking cessation information they need in an intuitive and efficient manner. This allows us to offer highly accessible smoking cessation support services to smokers motivated to quit, while effectively delivering reliable information to the general public interested in the subject.

Furthermore, by revitalizing outdated services and enhancing their appeal, we expect to encourage more citizens to use the No-smoking Guide service. This will, in turn, increase the effectiveness of national smoking cessation policies and contribute significantly to the promotion of public health.

Launch year: 2024

Attribution

Case study courtesy of the National Information Society Agency, Republic of Korea, Apolitical's Content Partner. Sourced from Use Cases of Public AI Service, Vol.2, https://eng.nia.or.kr/site/nia_eng/ex/bbs/View.do?cbIdx=31975&bcIdx=29182&parentSeq=29211